大规模浮动车流数据并行地图匹配方法

A Parallel Map-Matching Approach for Large Volume Floating Car Stream Data

提出了一种并行地图匹配方法,高效处理海量浮动车流数据。该方法顾及交通网络拓扑,指出网格过滤、距离过滤和方向过滤等策略减少邻近候选节点的数量,利用预先生成的最短路径列表减少最短路径计算量。基于非关系型分布式数据库实现了高效率的浮动车流数据并行地图匹配,利用武汉市的浮动车流数据进行了实验。实验结果表明,本文方法正确率为90.6%,计算效率能满足大规模浮动车流数据实时处理的需要。

Mp-matching floating car data is a fundamental task in traffic surveillance, traffic anomaly detection, and urban dynamic analysis. This study proposes a parallel map-matching approach to process streaming large volume floating car data. Considering the connectivity of a transportation net- work, the matching candidates are limited with a coarse spatial grid. A distance filter and a direction filter are combined to reduce the number of matching candidates. The trajectory between consecutive nodes is recovered with a shortest path list. The shortest path list in memory was developed to reduce the computation and speed up the matching process. A non-relational distributed database parallelizes the map-matching procedure. The performance of the presented approach was tested with large vol- ume floating car data in Wuhan, China. It demonstrates that this method achieves 90.62% correct map-matching results. This efficiency could fulfill the needs of real-time traffic monitoring, and will benefit trajectory analysis.